This prospective study investigates the causes of relapse following treatment for heroin addiction. Using a structured interview schedule, we propose to interview opiate addicts at three points in time: upon entry into treatment, after 6 months, and one year. We shall use multivariate statistical analysis (path regression analysis) to make causal inferences from the study's nonexperimental data. The analysis seeks to estimate a causal model of relapse that incorporates ideas from the writing of Wikler, Winick, Lindesmith, Brill, Waldorf, and McAuliffe. A causal model of relapse should suggest ways of reducing relapse rates among addicts in treatment.